Management of the acceptance degree for a technology transfer in automation field.

Omrani, Hichem ; Popescu, Catalin ; Boussier, Jean Marie 等

Abstract: For technology transfer to be successful in automation field, it has to suit the culture of the end user. Managerial and organizational methods adopted in industrialized countries may face serious problems when applied in developing countries. The way people think, feel and react is a result of their traditional ideas and their attached values such as comfort, worries for losing their job. An approach based on the belief and utility theory proposes to quantify an indicator of acceptance degree that could help the manager to focus his/her dissemination work on pertinent criteria affecting the technology transfer and to follow its evolution before and during the implementation of the project.

Key words: acceptance degree, transfer technology, belief theory, utility theory

1. A PROBLEM OF TECHNOLOGY TRANSFER

The different human, cultural, social and behavioural patterns of the people in developing countries demand that management systems of industrialized countries require prior adjustment or adaptation before they are transferred. Implementation of cultural calibration, however, is difficult due to the lack of relevant information on ethnic or cultural variability and the rather limited role that the Human Factors Engineering specialist is able to play during detailed engineering and short project time scales.

Several works exist, but they are generally focused on efficiency or effectiveness of implementation of an automation process (Hendrikse & McKinney, 2000). Studies of end user participation are limited or brought very late to the project with a simple evaluation for satisfaction that gives an indication of the acceptance of handling a device (Meister, 1985), without possibility to manage it.

However, define and quantify an indicator illustrating effects of several criteria causes many problems: how to take into account heterogeneous perceptions of criteria effects, which are the pertinent criteria, how to manage the indecision.

2. AVERAGE UTILITY OF HUMAN CRITERIA

Let us take an example of one criterion like "worry to lose a job" related to the installation of an automated chain in a company situated in a developing country (such as Romania). Let [OMEGA]={[H.sub.i],i=1, ..., p} be the domain of criteria levels which represent a finite set of mutually exclusive and exhaustive hypotheses, called the frame of discernment. In our application [OMEGA] = {Absolutely worried, Partly worried, Partly reassured, Absolutely reassured}={A.W, P.W, P.R, A.R}. A questionnaire is submitted to workers concerned by these changes and responses can be as singletons (one of levels) or as disjunctions (between two successive levels). By applying belief theory (Demspter, 1968), the indecision (frequently observed before implementation of a project) or the ignorance can be taken into account. Mass assignment is done by using frequency analysis (Denoeux, 2006), as is shown in figure 1.

[FIGURE 1 OMITTED]

Beliefs can be transformed into a probability measure denoted by BetP (Smets, 1994):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

where [absolute value of B] denotes the number of elements in the set B and m([empty set]) is the mass allowed to the conflict between workers' opinions. The results provided to the project manager are in numerical form starting from the computation of a utility ([u.sub.i]) for the criterion ([C.sub.i]):

[u.sub.i] = [P.summation over (k=1)]u([H.sub.k] x BetP([H.sub.k) (2)

where: u([H.sub.k]) is the utility of an evaluation level [H.sub.k], u([H.sub.k+1])[greater than or equal to]u([H.sub.k]) if [H.sub.k+1] is preferred to [H.sub.k], and BetP([H.sub.k]) is the "pignistic" probability related to [H.sub.k]. By using a linear function on the evaluation levels (e.g. u(A.W, P.W, P.W or P.R, P.R, A.R)=(0.25, 0.5, 0.75,1)), we obtain an utility (i.e. score) for "worry to lose the job" equal to 0.57. We will show in section 4 how this utility, which is an average value taking into account all workers responses will be exploited for acceptance level quantification.

3. CRITERIA CLASSIFICATION

The method for estimating criteria weights is based on the judgment of the evaluators and the belief theory. Let {[W.sub.i], i=1, ..., p} represent a group of workers and {[C.sub.k], k=1,..,n} be a group of criteria whose weights we want to determine {[[omega].sub.k], k=1, ..., n}. In our application the set of 'pertinence degree' was defined using 4 levels. The estimation of the criteria weights is given by the following equation:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)

where: i[member of] [OMEGA]={1,2,3,4}={Not Pertinent, Less Pertinent, Pertinent, Very Pertinent} is the set of pertinent degrees; Freq(i): frequency of appearance of pertinence degrees i; [absolute value of j]: cardinality of the set j, V (i) is level of degree of importance with a linear analytic form:

V (i) = (i-1)/(m-1) (4)

with m levels of pertinence and i [member of] [1;m]

After weights computation (equation 1), normalisation is necessary. Now, we consider that two criteria are tested: "worry to lose a job" (J) and "time efficiency at work" (E). In our application the pertinence degree of the criteria was evaluate by 6 workers (noted by [W.sub.i]). If a criterion is unknown, a respondent can reply by UNK "I do not know" (Table 1).

After normalisation, equation (3) becomes: [[omega].sub.J]=0.56 and [[omega].sub.E] = 0.44. These weights allow a manager to have an "average" classification of criteria that could affect the acceptance degree and to do a complementary work to correct the most important (in our case, the "worry to lose their job").

4. ACCEPTANCE DEGREE

We considered that the indicator for acceptance degree must illustrate the effects of all criteria tested by the manager. Generally a simple method of aggregation is applied for it which is based on the multi-attribute utility theory (MAUT) techniques. However, utility function is not necessary additive because the criteria set can be in interaction (in synergy or in redundancy). For it, we adapted the criteria aggregation by the 2-additive form of the Choquet integral (Rico, 2002) where global utility is:

u = [n.summation over (i=1)][[omega].sub.i] x [u.sub.i] - 1/2 [n.summation over (s=1)][n.summation over (t=s+1)] (5) [I.sub.st] x [absolute value of [u.sub.s]-[u.sub.t]] (5)

with: [[omega].sub.i] is the weight of the criterion [C.sub.i]; [u.sub.i] is the average value of utility defined by equation (3) and [I.sub.st] is the interaction between the criteria Cs and Ct which must be must computed.

For it a recent work (Grabisch & Perny, 2003) has proposed the concept of k-additive measure and some methods for computing the weights of interactions between criteria. These methods deal with classification problems (i.e. candidates ranking etc.) and they are not adequate in our work.

Our method is based on the belief theory. Let {[E.sub.i], i=1, ..., p} be a set of workers and {[C.sub.k], k=1,..,n} be a set of criteria whose index interactions we want to determine {[I.sub.ij], i,j=1, ..., n}. We define a set of levels for importance of interactions that the workers use for giving their opinion. In our application, this set contains 5 levels as follow: [OMEGA]={High Negative Interaction (-2), Less positive Interaction (-1), Not Interaction (0), Less positive Interaction (+1), High positive Interaction (+2)}.

Weights of interactions are computed as follow:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)

where k' [member of][OMEGA]; Freq(k') is frequency of appearance of levels for importance of interaction; [absolute value of l]: cardinality of the set l; BetP(k') is the "pignistic probability". V(k') reflects the importance of an interaction and it is given by the following equation:

V(k') = 2/(n=1) x k' (7)

For three interactions studied related to the criteria ("Worry to lose a job" (J.), Security level (S.), "Time efficiency at work"(E.)), the frequency of apparition is done in Table 2.

Interaction indexes are respectively: {J., S.}=0; {J, E.}=-0,875; {S., E.}= 0,625. Two pairs of criteria are strongly correlated but the effect of interaction "worry to lose his job" and "time efficiency at work" is the most important and it is perceived as harmful by the end users. By applying Equation 5, a value of acceptance degree for a phase of implementation of the project can be computed. This operation has an interest only if a comparative approach is designed. Suppose that this indicator is evaluated for several phases (before and during the implementation of the automation chain); average value of utility associated to each criterion could change, as well as weights of interactions perceived by end users. A manager could have a realistic idea of global perception of impacts of the project (using acceptance degree) and can imagine corrective strategies to manage a successful transfer technology (using weights of criteria and of interactions).

5. CONCLUSIONS AND PERSPECTIVES

Power and hierarchical decision-making, individual versus collective practices, and the perceived consequences may require consideration when technology is transferred. We have proposed a methodology for the evaluation of acceptance degree for a technology transfer in automation field, under the framework of utility and belief theory. Its interest is the capacity to combine linguistic evaluations of end users in an effective way, taking into account indecision, ignorance and suspicion. The manager is able to estimate pertinent criteria, evolution of acceptance degree and to establish corrective strategies during the implementation of the project.

This approach will be tested in Romanian companies, especially in automobile field where people attitudes concerning automation processes have been strongly affected during last few years, due to the success of transfer technology.

6. REFERENCES

Dempster, A.P. (1968). A generalisation of Bayesian inference, Journal of the Royal Statistical Society, 205-247.

Denoeux, T. (2006). Constructing Belief Functions from Sample Data Using Multinomial Confidence Regions. International Journal of Approximate Reasoning, Vol. 42, Issue 3, Pages 228-252, 2006.

Grabitch, M. & Perny, P.(2003). Agregation multicritere (Multicriteria aggregation). B. Bouchon-Meunier, C. Marsala (eds). Logique floue, principes, aide a la decision (Fuzzy logic, principles, aiding decision-making), Paris, Hermes, pp.81-120

Hendrikse, J. & McKinney, A., (2000). Human Factors Engineering and Cultural Calibration for an Offshore Platform Design, Conference RINA, London, England. Meister, D. (1985) Behavioural Analysis and Measures, John Wiley and Sons, New York

Rico, A. (2002). Modelisation des preferences pour l'aide a la decision par l'integrale de Sugeno (Preference modelling for decision-making aiding by Sugeno integral), Universite Paris I, these de doctorat (thesis).

Smets, P. & Kennes, R. (1994). The Transferable Belief Model. Artificial Intelligence, 66:191-243. Workers

Table 1: Opinions of workers

Workers/Criteria [W.sub.1] [W.sub.2] [W.sub.3]

J 1 2 3
E 1 {1,2} 3

Workers/Criteria [W.sub.4] [W.sub.5] [W.sub.6]

J 3 3 UNK
E 3 3 1

Table 2: Frequency of apparition

Levels/ -2 1 0 1 2 -2,- 1,2 UNK
Criteria

{J., S.} 0 0 0.5 0 0 0 0 0.5
{J., E} 0.5 0 0 0 0 0.5 0 0
{S, E} 0 0 0 0.5 0 0 0.5 0